Avoid btree Index() call on pattern with separators.

To avoid hard Index()'ing of given text with btree matcher we implement
an prefix_any and suffix_any matchers that can work well in many cases.

BTree matcher Index() will be implemented in upcoming commits to prevent
same bugs.

Fixes gobwas#23

compiler/compiler.go

@@ -43,37 +43,43 @@ func optimizeMatcher(matcher match.Matcher) match.Matcher {
 			return m
 		}
 
-		leftNil := m.Left == nil
-		rightNil := m.Right == nil
-
+		var (
+			leftNil  = m.Left == nil
+			rightNil = m.Right == nil
+		)
 		if leftNil && rightNil {
 			return match.NewText(r.Str)
 		}
 
 		_, leftSuper := m.Left.(match.Super)
 		lp, leftPrefix := m.Left.(match.Prefix)
+		la, leftAny := m.Left.(match.Any)
 
 		_, rightSuper := m.Right.(match.Super)
 		rs, rightSuffix := m.Right.(match.Suffix)
+		ra, rightAny := m.Right.(match.Any)
 
-		if leftSuper && rightSuper {
+		switch {
+		case leftSuper && rightSuper:
 			return match.NewContains(r.Str, false)
-		}
 
-		if leftSuper && rightNil {
+		case leftSuper && rightNil:
 			return match.NewSuffix(r.Str)
-		}
 
-		if rightSuper && leftNil {
+		case rightSuper && leftNil:
 			return match.NewPrefix(r.Str)
-		}
 
-		if leftNil && rightSuffix {
+		case leftNil && rightSuffix:
 			return match.NewPrefixSuffix(r.Str, rs.Suffix)
-		}
 
-		if rightNil && leftPrefix {
+		case rightNil && leftPrefix:
 			return match.NewPrefixSuffix(lp.Prefix, r.Str)
+
+		case rightNil && leftAny:
+			return match.NewSuffixAny(r.Str, la.Separators)
+
+		case leftNil && rightAny:
+			return match.NewPrefixAny(r.Str, ra.Separators)
 		}
 
 		return m

glob_test.go

@@ -120,6 +120,16 @@ func TestGlob(t *testing.T) {
 		glob(true, "/{rate,[0-9][0-9][0-9]}*", "/rate"),
 		glob(true, "/{rate,[a-z][a-z][a-z]}*", "/usd"),
 
+		glob(true, "{*.google.*,*.yandex.*}", "www.google.com", '.'),
+		glob(true, "{*.google.*,*.yandex.*}", "www.yandex.com", '.'),
+		glob(false, "{*.google.*,*.yandex.*}", "yandex.com", '.'),
+		glob(false, "{*.google.*,*.yandex.*}", "google.com", '.'),
+
+		glob(true, "{*.google.*,yandex.*}", "www.google.com", '.'),
+		glob(true, "{*.google.*,yandex.*}", "yandex.com", '.'),
+		glob(false, "{*.google.*,yandex.*}", "www.yandex.com", '.'),
+		glob(false, "{*.google.*,yandex.*}", "google.com", '.'),
+
 		glob(true, pattern_all, fixture_all_match),
 		glob(false, pattern_all, fixture_all_mismatch),
 
@@ -149,16 +159,16 @@ func TestGlob(t *testing.T) {
 		glob(true, pattern_prefix_suffix, fixture_prefix_suffix_match),
 		glob(false, pattern_prefix_suffix, fixture_prefix_suffix_mismatch),
 	} {
-		g, err := Compile(test.pattern, test.delimiters...)
-		if err != nil {
-			t.Errorf("parsing pattern %q error: %s", test.pattern, err)
-			continue
-		}
-
-		result := g.Match(test.match)
-		if result != test.should {
-			t.Errorf("pattern %q matching %q should be %v but got %v\n%s", test.pattern, test.match, test.should, result, g)
-		}
+		t.Run("", func(t *testing.T) {
+			g := MustCompile(test.pattern, test.delimiters...)
+			result := g.Match(test.match)
+			if result != test.should {
+				t.Errorf(
+					"pattern %q matching %q should be %v but got %v\n%s",
+					test.pattern, test.match, test.should, result, g,
+				)
+			}
+		})
 	}
 }
 

match/any_of.go

@@ -1,8 +1,6 @@
 package match
 
-import (
-	"fmt"
-)
+import "fmt"
 
 type AnyOf struct {
 	Matchers Matchers

match/prefix_any.go

@@ -0,0 +1,55 @@
+package match
+
+import (
+	"fmt"
+	"strings"
+	"unicode/utf8"
+
+	sutil "github.com/gobwas/glob/util/strings"
+)
+
+type PrefixAny struct {
+	Prefix     string
+	Separators []rune
+}
+
+func NewPrefixAny(s string, sep []rune) PrefixAny {
+	return PrefixAny{s, sep}
+}
+
+func (self PrefixAny) Index(s string) (int, []int) {
+	idx := strings.Index(s, self.Prefix)
+	if idx == -1 {
+		return -1, nil
+	}
+
+	n := len(self.Prefix)
+	sub := s[idx+n:]
+	i := sutil.IndexAnyRunes(sub, self.Separators)
+	if i > -1 {
+		sub = sub[:i]
+	}
+
+	seg := acquireSegments(len(sub) + 1)
+	seg = append(seg, n)
+	for i, r := range sub {
+		seg = append(seg, n+i+utf8.RuneLen(r))
+	}
+
+	return idx, seg
+}
+
+func (self PrefixAny) Len() int {
+	return lenNo
+}
+
+func (self PrefixAny) Match(s string) bool {
+	if !strings.HasPrefix(s, self.Prefix) {
+		return false
+	}
+	return sutil.IndexAnyRunes(s[len(self.Prefix):], self.Separators) == -1
+}
+
+func (self PrefixAny) String() string {
+	return fmt.Sprintf("<prefix_any:%s![%s]>", self.Prefix, string(self.Separators))
+}

match/prefix_any_test.go

@@ -0,0 +1,47 @@
+package match
+
+import (
+	"reflect"
+	"testing"
+)
+
+func TestPrefixAnyIndex(t *testing.T) {
+	for id, test := range []struct {
+		prefix     string
+		separators []rune
+		fixture    string
+		index      int
+		segments   []int
+	}{
+		{
+			"ab",
+			[]rune{'.'},
+			"ab",
+			0,
+			[]int{2},
+		},
+		{
+			"ab",
+			[]rune{'.'},
+			"abc",
+			0,
+			[]int{2, 3},
+		},
+		{
+			"ab",
+			[]rune{'.'},
+			"qw.abcd.efg",
+			3,
+			[]int{2, 3, 4},
+		},
+	} {
+		p := NewPrefixAny(test.prefix, test.separators)
+		index, segments := p.Index(test.fixture)
+		if index != test.index {
+			t.Errorf("#%d unexpected index: exp: %d, act: %d", id, test.index, index)
+		}
+		if !reflect.DeepEqual(segments, test.segments) {
+			t.Errorf("#%d unexpected segments: exp: %v, act: %v", id, test.segments, segments)
+		}
+	}
+}

match/suffix_any.go

@@ -0,0 +1,43 @@
+package match
+
+import (
+	"fmt"
+	"strings"
+
+	sutil "github.com/gobwas/glob/util/strings"
+)
+
+type SuffixAny struct {
+	Suffix     string
+	Separators []rune
+}
+
+func NewSuffixAny(s string, sep []rune) SuffixAny {
+	return SuffixAny{s, sep}
+}
+
+func (self SuffixAny) Index(s string) (int, []int) {
+	idx := strings.Index(s, self.Suffix)
+	if idx == -1 {
+		return -1, nil
+	}
+
+	i := sutil.LastIndexAnyRunes(s[:idx], self.Separators) + 1
+
+	return i, []int{idx + len(self.Suffix) - i}
+}
+
+func (self SuffixAny) Len() int {
+	return lenNo
+}
+
+func (self SuffixAny) Match(s string) bool {
+	if !strings.HasSuffix(s, self.Suffix) {
+		return false
+	}
+	return sutil.IndexAnyRunes(s[:len(s)-len(self.Suffix)], self.Separators) == -1
+}
+
+func (self SuffixAny) String() string {
+	return fmt.Sprintf("<suffix_any:![%s]%s>", string(self.Separators), self.Suffix)
+}

match/suffix_any_test.go

@@ -0,0 +1,47 @@
+package match
+
+import (
+	"reflect"
+	"testing"
+)
+
+func TestSuffixAnyIndex(t *testing.T) {
+	for id, test := range []struct {
+		suffix     string
+		separators []rune
+		fixture    string
+		index      int
+		segments   []int
+	}{
+		{
+			"ab",
+			[]rune{'.'},
+			"ab",
+			0,
+			[]int{2},
+		},
+		{
+			"ab",
+			[]rune{'.'},
+			"cab",
+			0,
+			[]int{3},
+		},
+		{
+			"ab",
+			[]rune{'.'},
+			"qw.cdab.efg",
+			3,
+			[]int{4},
+		},
+	} {
+		p := NewSuffixAny(test.suffix, test.separators)
+		index, segments := p.Index(test.fixture)
+		if index != test.index {
+			t.Errorf("#%d unexpected index: exp: %d, act: %d", id, test.index, index)
+		}
+		if !reflect.DeepEqual(segments, test.segments) {
+			t.Errorf("#%d unexpected segments: exp: %v, act: %v", id, test.segments, segments)
+		}
+	}
+}

util/strings/strings.go

@@ -1,6 +1,9 @@
 package strings
 
-import "strings"
+import (
+	"strings"
+	"unicode/utf8"
+)
 
 func IndexAnyRunes(s string, rs []rune) int {
 	for _, r := range rs {
@@ -11,3 +14,26 @@ func IndexAnyRunes(s string, rs []rune) int {
 
 	return -1
 }
+
+func LastIndexAnyRunes(s string, rs []rune) int {
+	for _, r := range rs {
+		i := -1
+		if 0 <= r && r < utf8.RuneSelf {
+			i = strings.LastIndexByte(s, byte(r))
+		} else {
+			sub := s
+			for len(sub) > 0 {
+				j := strings.IndexRune(s, r)
+				if j == -1 {
+					break
+				}
+				i = j
+				sub = sub[i+1:]
+			}
+		}
+		if i != -1 {
+			return i
+		}
+	}
+	return -1
+}